Assimilating Remote Sensing Phenological Information into the WOFOST Model for Rice Growth Simulation
Assimilating Remote Sensing Phenological Information into the WOFOST Model for Rice Growth Simulation
复制标题
DOI：
10.3390/rs11030268
复制
发表时间：
2019-01
期刊：
影响因子：
--
通讯作者：
G. Zhou;Xiangnan Liu;Ming Liu
中科院分区：
文献类型：
--
作者：
G. Zhou;Xiangnan Liu;Ming Liu
作物生长的精确模拟对产量估算、农田管理和气候变化至关重要。虽然作物模型和遥感数据的同化已经应用于作物生长模拟，但很少有研究考虑在物候方面优化作物模型。为了提高区域尺度上水稻生长模拟的精度，我们将中分辨率成像光谱仪（MODIS）时间序列数据中的物候信息同化到世界粮食研究（WOFOST）模型中。采用粒子群优化算法（PSO）优化WOFOST模型的物候初始发育阶段（IDVS）和移植日期（TD），使抽穗期和成熟期等物候模拟值与观测值之间的差异最小。物候同化提高了水稻生长模拟的准确性，3个野外数据的相关系数(R)分别为0.793、0822和0.813。该策略与增强型植被指数（EVI）时间序列同化策略性能相当，但计算时间更短。此外，实验结果证实了所提出的策略可以应用于不同空间分辨率的图像，并且在三个实验区模拟的LAImean差异小于0.35。本研究为模拟作物生长提供了一种高效、可扩展的物候发育过程同化策略。
Precise simulation of crop growth is crucial to yield estimation, agricultural field management, and climate change. Although assimilation of crop model and remote sensing data has been applied in crop growth simulation, few studies have considered optimizing the crop model with respect to phenology. In this study, we assimilated phenological information obtained from Moderate Resolution Imaging Spectroradiometer (MODIS) time series data into the World Food Study (WOFOST) model to improve the accuracy of rice growth simulation at the regional scale. The particle swarm optimization (PSO) algorithm was implemented to optimize the initial phenology development stage (IDVS) and transplanting date (TD) in the WOFOST model by minimizing the difference between simulated and observed phenology, including heading and maturity date. Assimilating phenology improved the accuracy of the rice growth simulation, with correlation coefficients (R) equal to 0.793, 0822, and 0.813 at three fieldwork dates. The performance of the proposed strategy is comparable with that of the enhanced vegetation index (EVI) time series assimilation strategy, with less computation time. Additionally, the result confirms that the proposed strategy could be applied with different spatial resolution images and the difference of simulated LAImean is less than 0.35 in three experimental areas. This study offers a novel assimilation strategy with regard to the phenology development process, which is efficient and scalable for crop growth simulation.